Ever since they tried to force me to upgrade from 10.4 to 10.5 on a macbook pro all because I wanted to do some iPhone development. I did not have the time to upgrade the machine at that time; So don't make me upgrade. That is when I discovered Fedora.
It is only a matter of time before this company crashes and burns... again. Can't employ SJ this time around to save it.
Sure. You can quote whatever numbers you like. CPU Cores: 103, 680 GPU Cores: 129,024
Total machine is 2PetaFlop and the GPUs contribute less than 10% of that.
Look, I realise that for the right job, the GPU is superior. But it is not anywhere near what we are being led to believe (again, according to the marketing material). I sure the people commissioning this machine knew what they were doing and what they needed was raw x86_64 grunt.
I'm not interested in a single program's performance on the GPU. I'm interested in the average use case. Specific cases will always require specific hardware. Even then, I don't believe it until I see the source code of each. Techniques people use on the GPU are somehow not used when it comes to the CPU.
GPUs: "We must use single stride arrays." CPUs: "We must use templated codes where memory is fragmented on the heap cause stl performs error checks."
Interesting to note that they didn't bother with too many gpu nodes. Reflects what we see with our users despite the abundance of marketing material from Nvidia.
Thanks for that, CSIRO troll. There is a difference between research grants and projects you retard.
I'm talking about the product development projects; The ones where other companies had existing contracts and the CSIRO came in and said 'No we can do better. We are the CSIRO.' Years later on delivery date and no product. Just a little document that stated how hard it was and there were all sorts of difficulties. "But if you give us another couple of million, we'll deliver it. We promise."
The government, universities and some private industry are wising up.
Even the scientific board in charge of deciding where the SKA will be located are onto the half-truths of the CSIRO. Once again, the CSIRO are out in force, slagging off SA for 'political and instability' reasons instead of focusing on the technical reasons (the arguments for which they have already lost). Now that it is looking like SA will get the SKA and the CSIRO are now fighting dirty.
I've been meeting them for quite some years now and the CSIRO guys I've met are about protectionism. Since they lost their.edu status, they are about turning a buck. They will lie, cheat and steal their way through any bit of technology and pawn it off as their own. They may once have had skill but those days have gone.
The CSIRO are pretending to be elite. They plant themselves into the University system and pinch any idea that has the smallest amount of creativity. They will take established conferences and hijack them as their own. I bet you they had meetings about the recent attention of the WIFI thing and thought about how they can try and keep the momentum going. You wouldn't believe the extent they will advertise because they know that this is how you attract research and development money in Australia. Glossy mags and smiling pictures with MPs and popular projects as backdrops are taking research dollars from.edu and into putting it into.com. I wouldn't mind so much except they never deliver on more than 90% of projects. Just Word documents.
Sure, there may be a couple of greybeards that still create, but none of the new guys do. But when the chief scientist, ceo type, publicly states that the future of energy is fossil fuels and not renewable energies, something is very, very wrong. That was until it was popular to be green.
Sorry mate, but sounds like you're making excuses.
I ride 20km each way to work, Monday to Friday. 200km weekly. You do not need workouts after work with that sort of regime.
I leave clothes (inc. shoes) at work. They don't come home (take to dry cleaners). I arrive at work about 1 hour before everyone else does so I have cooled down by then. There is also this invention called anti-perspirant.
I do this in Brisbane Australia. Coming out of 30-36 Celsius temps now and humidity comparable to yours. I actually prefer to ride in these conditions. Riding in the rain is enjoyable.
After Chapman's death, speculation of a Python revival inevitably faded. Idle stated: "We would only do a reunion if Chapman came back from the dead. So we're negotiating with his agent."
You also share housing, eat very cheaply, use those small coins, walk everywhere or ride that bike you found, and the hundreds of other ways to do things on the cheap (and when you're young).
Admittedly, in Australia, the education is not as expensive as the US, but it is quite common to work your way through and graduate with zero loans. It is still.
You mention GPU but can you use get the solution up and running as quickly as the cpu solution? Optimised multi-gpu solutions are not that easy as the programmer has to do all the heavy lifting.
Does the code vectorise? If is does, then I'd be tempted to go with as many dual socket Intel machines as you can. Are you able to use the Intel compiler (leveraging into the MKL, IPP and IMF as much as possible). This assumes that communication is low. You are not going to have the cash for a low latency, high speed interconnect. The Intel compiler is free for linux and non-commercial use.
If the code doesn't vectorise, then I'd go for the recent quad socket, 16 core AMD setups and just go for blind horsepower.
Do you work for an organisation that already has big compute requirements and a system in place? Can you buy time there? Our installation runs at about a quarter the price of EC2 (and that includes people for installing software, configuring environments and providing compiler licenses). Can you contribute money to this group? $15K with us would get you 4 nodes of dual socket, 6 core Xeons, dual lane 256 bit wide registers and 2.8GHz clock.... 1.075TFlop. And access to the compilers to get very close to that theoretical performance easily.
There are all sorts of things that can steer the direction is any of the above and even others. Good luck.
Slashdot Mirror
May as well grab the devel package as well. One can only hope some development occurs.
apt-get install functioning-brain-cell functioning-brain-cell-devel
Fuck I hate Apple.
Ever since they tried to force me to upgrade from 10.4 to 10.5 on a macbook pro all because I wanted to do some iPhone development. I did not have the time to upgrade the machine at that time; So don't make me upgrade. That is when I discovered Fedora.
It is only a matter of time before this company crashes and burns ... again. Can't employ SJ this time around to save it.
It was nice knowing you.
Any power consumption data available yet?
Too right! For years I've been "campaigning" for a drug Olympics. Let's really see what human bodies are capable of.
Your parentheses are unbalanced.
Sure. You can quote whatever numbers you like.
CPU Cores: 103, 680
GPU Cores: 129,024
Total machine is 2PetaFlop and the GPUs contribute less than 10% of that.
Look, I realise that for the right job, the GPU is superior. But it is not anywhere near what we are being led to believe (again, according to the marketing material). I sure the people commissioning this machine knew what they were doing and what they needed was raw x86_64 grunt.
I'm not interested in a single program's performance on the GPU. I'm interested in the average use case. Specific cases will always require specific hardware. Even then, I don't believe it until I see the source code of each. Techniques people use on the GPU are somehow not used when it comes to the CPU.
GPUs: "We must use single stride arrays."
CPUs: "We must use templated codes where memory is fragmented on the heap cause stl performs error checks."
I'm sick of it man. Sick of it :-)
Interesting to note that they didn't bother with too many gpu nodes. Reflects what we see with our users despite the abundance of marketing material from Nvidia.
5040 'standard' compute nodes: dual E5-2680 processors; 64GB RAM
360 'bulk' compute nodes: quad EX-X7560; 128GB RAM
144 GPU nodes: dual M2050
Another 90 'super' nodes on order: 128core, 512GB RAM
Cores: 103,680
GPUs: 288
Almost token GPU offering. These guys must do real work on it.
Why should Americans pay for Internet research so foreigners can use it for free? I'm Australian and your argument is useless.
If it weren't for the Americans, there would be no Internet to try and make wireless. Last time I checked, I don't pay a license for the Internet.
Wireless was not new research. It was built on the shoulders of giants.
s/you're/your/g
I said troll. Not patent troll. You're reading comprehension skills are something to be desired mate.
Thanks for that, CSIRO troll. There is a difference between research grants and projects you retard.
I'm talking about the product development projects; The ones where other companies had existing contracts and the CSIRO came in and said 'No we can do better. We are the CSIRO.' Years later on delivery date and no product. Just a little document that stated how hard it was and there were all sorts of difficulties. "But if you give us another couple of million, we'll deliver it. We promise."
The government, universities and some private industry are wising up.
Even the scientific board in charge of deciding where the SKA will be located are onto the half-truths of the CSIRO. Once again, the CSIRO are out in force, slagging off SA for 'political and instability' reasons instead of focusing on the technical reasons (the arguments for which they have already lost). Now that it is looking like SA will get the SKA and the CSIRO are now fighting dirty.
It is a national embarrassment and money pit.
No. CSIRO are trolls.
I've been meeting them for quite some years now and the CSIRO guys I've met are about protectionism. Since they lost their .edu status, they are about turning a buck. They will lie, cheat and steal their way through any bit of technology and pawn it off as their own. They may once have had skill but those days have gone.
The CSIRO are pretending to be elite. They plant themselves into the University system and pinch any idea that has the smallest amount of creativity. They will take established conferences and hijack them as their own. I bet you they had meetings about the recent attention of the WIFI thing and thought about how they can try and keep the momentum going. You wouldn't believe the extent they will advertise because they know that this is how you attract research and development money in Australia. Glossy mags and smiling pictures with MPs and popular projects as backdrops are taking research dollars from .edu and into putting it into .com. I wouldn't mind so much except they never deliver on more than 90% of projects. Just Word documents.
Sure, there may be a couple of greybeards that still create, but none of the new guys do. But when the chief scientist, ceo type, publicly states that the future of energy is fossil fuels and not renewable energies, something is very, very wrong. That was until it was popular to be green.
An event such as this?
Sorry mate, but sounds like you're making excuses.
I ride 20km each way to work, Monday to Friday. 200km weekly. You do not need workouts after work with that sort of regime.
I leave clothes (inc. shoes) at work. They don't come home (take to dry cleaners). I arrive at work about 1 hour before everyone else does so I have cooled down by then. There is also this invention called anti-perspirant.
I do this in Brisbane Australia. Coming out of 30-36 Celsius temps now and humidity comparable to yours. I actually prefer to ride in these conditions. Riding in the rain is enjoyable.
I find that an odd sequence of words from Knuth; "basically impossible".
Surreal.
After Chapman's death, speculation of a Python revival inevitably faded. Idle stated:
"We would only do a reunion if Chapman came back from the dead. So we're negotiating with his agent."
I wonder, whether in the future, the average length of songs will decrease? Song part A, part B and part C.
You also share housing, eat very cheaply, use those small coins, walk everywhere or ride that bike you found, and the hundreds of other ways to do things on the cheap (and when you're young).
Admittedly, in Australia, the education is not as expensive as the US, but it is quite common to work your way through and graduate with zero loans. It is still.
Three days, not four. Australia's Monday is Samoa's Sunday. Samoa's Friday is Australia's Saturday.
No longer. All days now align.
A work friend's response:
----------------
From what I guess, (just for fun)
In English,
1."Oh Dear"=="Oh God"
divide "Oh" on both sides=> dear==god
thus "dearbook" =="godbook"
In Chinese,
"tian"=="god"
"shu"=="book"
"tian shu" literally means a book that only God can read. It is basically a book has nothing but blank pages. :)
Hamilton's Night's Dawn Trilogy
You mention GPU but can you use get the solution up and running as quickly as the cpu solution? Optimised multi-gpu solutions are not that easy as the programmer has to do all the heavy lifting.
Does the code vectorise? If is does, then I'd be tempted to go with as many dual socket Intel machines as you can. Are you able to use the Intel compiler (leveraging into the MKL, IPP and IMF as much as possible). This assumes that communication is low. You are not going to have the cash for a low latency, high speed interconnect. The Intel compiler is free for linux and non-commercial use.
If the code doesn't vectorise, then I'd go for the recent quad socket, 16 core AMD setups and just go for blind horsepower.
Do you work for an organisation that already has big compute requirements and a system in place? Can you buy time there? Our installation runs at about a quarter the price of EC2 (and that includes people for installing software, configuring environments and providing compiler licenses). Can you contribute money to this group? $15K with us would get you 4 nodes of dual socket, 6 core Xeons, dual lane 256 bit wide registers and 2.8GHz clock .... 1.075TFlop. And access to the compilers to get very close to that theoretical performance easily.
There are all sorts of things that can steer the direction is any of the above and even others. Good luck.
For a very few specialist problems. Just like FPGAs.
On the whole, are they're useless for most applications requiring performance. A lot of people bought into the hype.
If the American government is not going to fund NASA (properly), can they at least put up a paypal account/donation thingy or something?
>$US600Billion for military
$US20Billion for NASA
Sorry, but as a foreigner I'm happy to throw a couple of regular bucks for a good cause.